arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2304.01933cs.CL
收藏

LLM-Adapters:面向大语言模型参数高效微调的适配器家族

LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models

Singapore University of Technology and Design(新加坡科技设计大学);Singapore Management University(新加坡管理大学);Southwest Jiaotong University(西南交通大学);DAMO Academy, Alibaba Group(阿里巴巴集团达摩院);University of Electronic Science and Technology of China(电子科技大学)

Zhiqiang Hu, Lei Wang, Yihuai Lan, Wanyu Xu, Ee-Peng Lim, Lidong Bing, Xing Xu, Soujanya Poria, Roy Ka-Wei Lee

AI总结:

本文提出集成多种适配器与开源大语言模型的易用框架LLM-Adapters,通过实验验证基于适配器的参数高效微调可让7B规模模型在两类推理零样本任务上达到甚至超过175B大模型的性能。

摘要:

大语言模型(LLM)(如GPT-4和ChatGPT)的成功推动了众多高性价比且易获取的替代模型的发展,这些模型通过使用特定任务数据(如ChatDoctor)或指令数据(如Alpaca)微调开源LLM得到。在各类微调方法中,基于适配器的参数高效微调(PEFT)无疑是最受关注的方向之一,因为它只需微调少量外部参数而非整个LLM,就能取得相当甚至更优的性能。为推动LLM的PEFT方法进一步研究,本文提出LLM-Adapters,这是一个易用的框架,可将多种适配器集成到LLM中,并能针对不同任务执行这些基于适配器的LLM PEFT方法。该框架涵盖LLaMA、BLOOM、GPT-J等当前最先进的开源LLM,以及串行适配器、并行适配器、基于提示的学习和基于重参数化的方法等常用适配器。此外,我们针对适配器类型、放置位置和超参数对每种基于适配器的方法最优设计的影响开展了大量实证研究。我们在来自算术推理和常识推理两类不同推理任务的14个数据集上评估了适配器的有效性。结果表明,在规模较小的LLM(7B)中使用基于适配器的PEFT,仅需少量额外可训练参数,在两类推理任务的零样本推理中,性能就能与强大的LLM(175B)相当,部分情况下甚至更优。

英文摘要:

The success of large language models (LLMs), like GPT-4 and ChatGPT, has led to the development of numerous cost-effective and accessible alternatives that are created by finetuning open-access LLMs with task-specific data (e.g., ChatDoctor) or instruction data (e.g., Alpaca). Among the various fine-tuning methods, adapter-based parameter-efficient fine-tuning (PEFT) is undoubtedly one of the most attractive topics, as it only requires fine-tuning a few external parameters instead of the entire LLMs while achieving comparable or even better performance. To enable further research on PEFT methods of LLMs, this paper presents LLM-Adapters, an easy-to-use framework that integrates various adapters into LLMs and can execute these adapter-based PEFT methods of LLMs for different tasks. The framework includes state-of-the-art open-access LLMs such as LLaMA, BLOOM, and GPT-J, as well as widely used adapters such as Series adapters, Parallel adapter, Prompt-based learning and Reparametrization-based methods. Moreover, we conduct extensive empirical studies on the impact of adapter types, placement locations, and hyper-parameters to the best design for each adapter-based methods. We evaluate the effectiveness of the adapters on fourteen datasets from two different reasoning tasks, Arithmetic Reasoning and Commonsense Reasoning. The results demonstrate that using adapter-based PEFT in smaller-scale LLMs (7B) with few extra trainable parameters yields comparable, and in some cases superior, performance to powerful LLMs (175B) in zero-shot inference on both reasoning tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑